Human Capital Intensity in Technology-Based Firms Located in Portugal: Do Foreign Multinationals Make a Difference?

Transcription

1 F E P W O R K I N G P A P E R S Research Work in Progress n. 187, August 2005 Human Capital Intensity in Technology-Based Firms Located in Portugal: Do Foreign Multinationals Make a Difference? Ana Teresa Tavares Aurora A. C. Teixeira CEMPRE - Centro de Estudos Macroeconómicos e Previsão Rua Dr. Roberto Frias, Porto Tel Faculdade de Economia da Universidade do Porto Tel

2 HUMAN CAPITAL INTENSITY IN TECHNOLOGY-BASED FIRMS LOCATED IN PORTUGAL: DO FOREIGN MULTINATIONALS MAKE A DIFFERENCE? Ana Teresa Tavares atavares@fep.up.pt Aurora A. C. Teixeira ateixeira@fep.up.pt CEMPRE, Faculdade de Economia, Universidade do Porto, Rua Dr. Roberto Frias, Porto Portugal ABSTRACT This paper contributes to the scarce empirical literature on the impact of foreign ownership on human capital intensity. New evidence is provided, based on a comprehensive, large-scale survey of technology-based firms located in Portugal. Using two alternatives measures of human capital (one based on skills, another on education), the key findings are that: (1) foreign ownership directly (and significantly) impacts on firms general human capital (education); (2) foreign ownership indirectly (and significantly) impacts on firms specific human capital (skills); (3) the total impact of foreign ownership on firms human capital intensity is higher for education- (general) than for skills- (specific) related human capital intensity. Other findings are that younger and smaller firms tend to be more human capital intensive, and that export patterns are not significantly related to human capital intensity. Giving the critical importance of both FDI and human capital development for an economy like Portugal (lagging behind in terms of human capital stock, and seeming to have lost part of its attractiveness as an FDI location), the paper discusses related policy implications. Keywords: Foreign direct investment (FDI), multinational enterprises (MNEs), human capital, education, technology-based firms (TBFs). JEL Classification: J24; F23. CEMPRE - Centro de Estudos Macroeconómicos e Previsão - is supported by the Fundação para a Ciência e a Tecnologia, Portugal, through the Programa Operacional Ciência, Tecnologia e Inovação (POCTI) of the Quadro Comunitário de Apoio III, which is financed by FEDER and Portuguese funds. 1

3 1. Introduction Human capital and foreign direct investment (FDI) are widely seen as key engines of economic growth and development (Romer, 1986; Lucas, 1988; Grossman and Helpman, 1991; Dunning, 1993). While there is considerable literature focusing on either FDI or human capital in isolation, the specific link between the two has been less researched. The issue has further interest given a potential two-way causality between human capital and FDI. Human capital has been recognised as an important FDI determinant (Noorbakhsh et al., 2001). In turn, foreign owned companies might be relevant contributors to human capital formation, as they affect both the demand and supply of skilled labour (Slaughter, 2002). Most extant work focuses on the first direction of impact. Studies highlighting the impact of FDI on human capital formation are scarce, rather exploratory (typically opinions and conceptual literature reviews) and mainly based on developing countries. 1 Even though there are very comprehensive and useful literature reviews (for instance, Blomström and Kokko, 2003; Rasiah, 2005; it is fair to say that empirical studies are very scarce. An exception is Narula and Marin (2003), a thorough empirical study comparing foreign-owned versus domestic firms in Argentina as regards the quantity and quality of human capital they employ, further linking that to technological spillovers. In this paper we propose to contribute to this scarce empirical literature on the relationship between human capital and FDI by investigating the relevance of foreign ownership for the human capital intensity of technology based firms (TBFs) located in Portugal. This study focuses on an under-researched empirical setting, the Portuguese case. For Portugal, no similar study exists, relating human capital intensity and foreign versus domestic ownership. Moreover, the themes of FDI and human capital development are particularly relevant to this peripheral European economy, marked by convergence difficulties vis-à-vis the European Union (EU), and with a considerable human capital and technological disadvantage relative to developed countries in general. In addition, Portugal embraced in the last years a proactive FDI attraction policy (with a recently opened investment promotion institution and accompanying legislation geared for that), recognising the potential role foreign multinationals could have in upgrading Portugal s industrial fabric and in the accumulation of competences. Therefore, the theme underlying this paper is a critical one for the Portuguese economy. We will test whether foreign-owned TBFs behave 1 Mainly stimulated by specific OECD and World Bank initiatives aiming to call the attention to the relevance of human capital development in countries with a modest human capital stock, or a dearth of initiatives to improve education and skills upgrading. The most comprehensive collection of papers, resulting from a technical meeting specifically aiming to debate FDI, human capital and education in developing countries, can be found in OECD (2001). 1

4 differently than their domestic counterparts in what respects human capital intensity. This is the research question here addressed. The remainder of the paper is structured as follows. Section 2 reviews extant literature on human capital and FDI, highlighting their connection with economic growth and development, and puts forward the hypotheses tested in the paper. Section 3 presents the data, providing the descriptive statistics of the respondent TBFs located in Portugal, specifically concerning their human capital traits and foreign ownership structure. The following section explains the empirical methodology, presents the econometric models estimated, and discusses the results obtained. The final section concludes, derives policy implications, and suggests avenues for future research. 2. Human capital, FDI and technology development Since the late 1980s, education (mainly at higher levels) became increasingly associated with economic performance and international competitiveness (e.g. Aldcroft, 1992). In particular, with the emergence of the so-called endogenous growth theories, an important role the engine of growth (Ehrlich, 1990) has been assigned to human capital. The development of both the Lucas (1988) approach (inspired by the work of Becker) and the work of Nelson- Phelps (1966) converge in a positive effect of educational attainment on the productivity of workers, and hence on growth. Nowadays, most economists and policy-makers consider human capital a key productive asset, highly complementary with technological capital. Earlier studies focusing on crosscountry growth performance over long periods generally yielded positive results (e.g., Barro and Lee, 1993). Although the amount of theoretical and empirical studies on human capital at firm or establishment level is already significant (Teixeira, 2002), most analyses are more aggregated, and mainly focusing on issues of economic growth (Wößmann, 2003) or rate-of-return analysis (Sianesi and van Reenen, 2003). The present analysis aims to contribute for a finer view of these issues, in addition relating human capital to multinationality, an underresearched area. FDI is nowadays one of the most topical issues and a particular focus of policy in many countries (Hanson, 2001; Young, 2004; Young and Tavares, 2004), owing to its sheer scale and importance, as well as that FDI is often seen by policy-makers (Portuguese included) as a fast track panacea for growth and development. Most countries (developed and developing 2

5 alike) scramble to attract FDI projects (Oxelheim and Ghauri, 2003), based on the common wisdom, or the stylised fact that multinationals bring positive externalities ( spillovers ) 2 to the domestic economy, stimulating development, growth, employment (quantity and skill upgrading/human capital development), wages, exports, technological and managerial innovation, productivity, domestic entrepreneurship and other impacts such as demonstration, agglomeration, competition and linkage effects, thereby enhancing macroeconomic conditions, the sophistication of the local industrial fabric and of the indigenous workforce. However, and even if this is the most common view, a growing body of literature questions the magnitude of these effects, or even if they are positive (for instance Haddad and Harrison, 1993; Aitken and Harrison, 1999; Kathuria, 2000; Castellani and Zanfei, 2003) as well as the economic efficiency (opportunity cost) of the sizeable incentives 3 given by countries through their proactive FDI attraction policies (Hanson, 2001; Oxelheim and Ghauri, 2003; Young, 2004). Just like research focusing on human capital, FDI impact issues are better researched at micro (firm) level (Görg and Strobl, 2001), further supporting the approach (as regards unit of analysis) used in this paper. Analysing now the links between human capital and FDI, the complementarity between both aspects has been investigated, in different ways. Quite often, the FDI-human capital link is treated as a secondary issue in studies focusing primarily on the FDI-growth nexus (Borensztein et al., 1998; Ram and Zhang, 2002). While the former study argued in favour of such complementarity, the second did not find evidence supporting that, emphasising the need for further research. Borensztein et al. (1998) refer another interesting relationship between FDI and human capital: when analysing the impact of FDI on economic growth, they find a positive link, but only provided that a minimum threshold of human capital exists [i.e., corroborating the absorptive capacity hypothesis (Cohen and Levinthal, 1989)]. Although this issue is not our focus here, we respond to their call for the importance of studying the effects of FDI on human capital. In line with Borensztein et al. s (1998) idea of a minimum threshold of human capital, both Xu (2000) and Saggi (2000) found that, in the absence of adequate human capital, spillovers (namely of a technological nature, and productivity spillovers) may simply be unfeasible. 2 There is a vast literature on spillovers and on the potential impact of multinational firms (Rodríguez-Clare, 1996; Blomström and Kokko, 1998; Markusen and Venables, 1999; for a review, see Görg and Strobl, 2001, and Tavares and Young, 2005). 3 Of particular interest for this paper is the fact that in Portugal (as well as in the other Cohesion countries) there have been vast sums awarded in the form of training grants (mainly funded by the European Social Fund), as well as other financial grants (provided by the European Regional Development Fund) directly financing FDI projects, and indirectly through funding relevant infrastructure such as business parks and roads. 3

6 Therefore, even when dealt with as a secondary research question, extant work reaches the conclusion that human capital is crucial to enable the spillovers that underpin economic growth. The two-way causality between FDI and human capital has been also studied in its own right, though we think that not often as much as the importance of the topic warranted. As Blomström and Kokko (2001) refer, the FDI and human capital relationship is complex and highly non-linear, and our knowledge in this vein is still sketchy. However, and considering firstly the direction of causality from human capital to FDI (i.e. human capital as a determinant of FDI inflows), this relationship has been more researched (Noorbakhsh et al., 2001). The main lacuna remains, thus, in the direction of causality from FDI to human capital, that is, does foreign-ownership matter as regards firms human capital intensity? As referred to above, Narula and Marin (2003) constitute an exception to this scarce empirical literature on the impact of FDI on firms human capital. They found that overall, foreign-owned firms hire more qualified workers, and have a more skilled workforce, than their local counterparts. In a relevant conceptual piece, Slaughter (2002) notes that MNEs can raise both the demand and the supply of labour (Slaughter, 2002). MNEs tend to be very technologically intensive firms (Slaughter, 2002), thereby demanding skilled staff to work along their knowledge-specific assets. This knowledge intensity can help raising host country demand for skills, through at least three main channels: technology transfer, technological spillovers to domestic firms, and capital investments (the latter for both foreign and domestically-owned firms alike). Saggi (2000) emphasises the (often neglected, in his words) importance of labour turnover as a channel for inter-firm technology diffusion (particularly, from MNEs to domestic firms). The interaction between the market for labour and for goods is no new, and in related literature (focusing on spillovers) it has been noted by Glass and Saggi (1999), and Saggi (2000). Therefore, the stimulus provided by MNEs, their demand patterns, their competitive pressure, and potential better practices in human resource development, would, ideally, spur a virtuous circle between foreign-owned and domestic firms, raising the bar of qualification and making the laggards converge hopefully towards the more advanced human capital hiring companies. It is worth analysing in more detail the mechanisms through which MNEs can contribute to increase the demand of human capital, an impact highlighted in many studies (Feenstra and Hanson, 1997; Markusen and Venables, 1997; Slaughter, 2002). That may occur via vertical or horizontal linkages. Concerning vertical linkages, if MNEs demand high 4

7 standards to suppliers and subcontractors, then the latter have to adjust to these requirements, and often have to hire highly skilled/qualified collaborators. This increase in the demand of skilled labour may happen also through horizontal linkages, for instance via demonstration effects and competitive pressure on firms in the same industry (foreign-owned or domestic), raising the bar again for all industry players. Now turning to the possibility of MNEs influencing the supply of skilled labour, that can happen both at the micro- and at the macro-level (Slaughter, 2002). This author (based on aggregate data) offers some empirical evidence about the existing correlation between FDI and skill upgrading, for 16 countries and from 1982 to MNEs can help increasing the supply of skilled labour by sponsoring education programmes (usually at the tertiary level), by formal training of their workforce (done inside or outside the firm), and by informal, onthe-job training. These initiatives can be an important source of technical and managerial skills for the workers (that can eventually spill over, especially if labour mobility to domestic firms occurs, or if entrepreneurial workers set up their own companies). The impact of MNEs on the labour market is often explored via the impact on wages (Feenstra and Hanson, 1997; Markusen and Venables, 1997; Bruno et al., 2004). The latter (and more recent) study, contrary to previous work, found that FDI did not affect the relative demand for skilled labour, in some Eastern European countries. Hence, the correlation between FDI and skill upgrading/superior human capital intensity is not uncontroversial, which is testified by the mixed results achieved by different studies. Nevertheless, on balance, we would go along with the dominant impression reported in previous work, that foreign owned firms tend to be more human capital intensive than domestic firms. Our first hypothesis is thus: Hypothesis 1: Foreign owned TBFs present higher human capital intensity than their nationally owned counterparts. Another crucial relationship, that has relevance to understand better the FDI-human capital link, is the FDI-technology nexus. There is a vast literature highlighting the central role of MNEs as producers and disseminators of knowledge and technological innovations (Teece, 1977; Cantwell, 1989; Narula, 2003). This literature has different types of focus, from studying the technology transfer from foreign MNEs to indigenous firms, to a more interactive technology creation process between the two types of agents, or even to investigate the intra-firm transfer and sharing of knowledge between parent firms and 5

8 subsidiaries (including reverse transfer from subsidiaries to parents/strategic asset-seeking). No matter the lens adopted, it is clear that human capital matters in all these dynamics. Without properly trained or educated human resources, these processes of technology/knowledge transfer and creation, reactive or proactive, cannot occur effectively. Therefore, several authors (e.g. Haddad and Harrison, 1993) noted how critical for technology dissemination is the absorptive capacity (Cohen and Levinthal, 1989, 1990) of the receiving, or partner firms (which is, we argue, in fact the absorptive capacity of the staff working in these firms). A related topic to knowledge production concerns R&D activities. A rich stream of literature has highlighted the strong connection between multinational enterprises and R&D (Ronstadt, 1977; Pearce, 1989; Kuemmerle, 1999). Caves (1996: 163) noted that the affinities between R&D and the MNE are numerous, considering that the extent of R&D spending constitutes an excellent predictor of MNE activity in an industry. A few of the largest global multinationals account for the lion s share in R&D in several industries (e.g. pharmaceuticals, chemical). R&D activities can be thus hypothesised to be correlated to foreign ownership. In turn, performing R&D operations requires highly qualified and skilled professionals. This leads us to put forward the following hypothesis: Hypothesis 2: The impact of foreign ownership on TBFs human capital intensity is higher the more intensive are TBFs R&D efforts. One of the outcomes of the globalisation process is the growth of alliances, cooperation in technology development, or Strategic Technology Partnering (Narula, 2003). One of the ways in which this Strategic Technology Partnering (STP) may be materialised is through linkages between firms and other knowledge-producing agents (such as universities and research laboratories). We would expect that foreign-owned TBFs, because they are technology-based and because they are foreign (thus incurring a cost or liability of foreignness [Hymer, 1960/76; Zaheer, 1995]), may profit from developing such linkages like contacts with universities. There is not, to the best of our knowledge, a well established and systematic literature specifically on multinationals and university linkages. There is, however, a considerable stream of research on firms (regardless of ownership) and university contacts (Laursen and Salter, 2004; Mowery and Sampat, 2004). In order for the firm-university relationship to be productive, the firm in question needs to have adequate human capital, i.e. 6

9 staff that is able to interact and understand the university partners (the latter are, by definition, very qualified). The consideration of all these arguments leads us to hypothesise that: Hypothesis 3: The impact of foreign ownership on TBFs human capital intensity is higher the more frequent are TBFs contacts with Universities. 3. Methodology 3.1. Data This analysis is based on data gathered through a questionnaire survey. The firms surveyed were drawn from the Markelink 2004 list, which comprises firms located in Portugal that declared and publicised R&D activities. 4 This was the best possible list of companies publicly available, in order to obtain a credible list of TBFs located in Portugal. There are two very reputed and comprehensive alternatives, notably the company lists used by the Community Innovation Survey (CIS) and Observatório para a Ciência e Ensino Superior (OCES) survey, but these are not publicly disclosed owing to statistical secrecy. The list of companies we use, Markelink, includes 703 companies, representing 85% of CIS III innovative firms and encompassing a much higher number of firms than those considered as innovative by the last OCES survey, therefore it is a representative list. It is important to note that, similarly to CIS and OCES, Markelink list encompasses firms from all industries located within the Portuguese territory (including Azores and Madeira islands), and, differently from CIS, covers all size classes. The questionnaire was sent in November-December 2004 to all firms listed in the Markelink list (703) plus 4 firms that we knew (through the available on-line OCES list of Portuguese firms with the largest R&D expenditures in 2001) that performed R&D activities. 5 By mid- December, 425 complete valid replies were received, which represents an effective response rate of almost 61%. This is a surprisingly high response rate for a non-compulsory survey. As Harzing (1997) noted, non compulsory industrial surveys are typically plagued by low response rates. For instance, CIS III inquiry, which is compulsory, the response rate was 45,8% in the case of Portugal (Bóia, 2003) and 41,7% for the U.K. (Stockdale, 2002). 6 4 Markelink is a private information supplier who manages several databases, covering a wide-ranging set of issues, including R&D ( 5 For a detailed description of the implementation of this survey see Costa and Teixeira (2005). 6 Using a formula for computing the size of the sample, in random samples, based on a pessimistic scenario (Vicente et al., 1996), a sample size of 425 observations (in a population of 697 firms the difference between the 707 surveyed and 10 that we found to be out of business) would lead, for a confidence level of 95%, to a precision of approximately

10 Therefore, the dataset gathered through this survey is remarkably comprehensive and representative of the relevant population of firms Variables and descriptive statistics Dependent variable Since we are interested in explaining the human capital intensity of TBFs, our dependent variable is the proportion of top skilled and top educated workers in total employment. Although skills and education are treated in countless studies as synonymous concepts (e.g., Harris and Helfat, 1997), more accurately they are distinct (though interrelated) concepts. Skills can be acquired through education and (formal) training but also (and mainly) through the course of people s activities at work (i.e., learning-by-doing). Rosen (1986) points to the fact that most specific job skills are learned from performing the work activities themselves. Formal schooling complements these skills, both by providing a body of general knowledge and principles for students, as well as teaching them how to learn (Teixeira, 2005). In order to capture both components of human capital we test human capital intensity by using these two alternative (though interrelated) ways of measuring it. This is reflected in the alternative model specifications presented later (Table 3). Firms were asked about the number of total workers and the number of workers with an engineering degree, which tend to represent a more firm-industry specific human capital component, and the number of workers with 12 or more years of schooling (post-secondary school), a more general component of human capital (Becker, 1962). Then we compute two widely used ratios for proxying the human capital intensity variable: (i) the number of top skilled workers over total employment, being top skills measured by the number of engineers (Wood and Ridao-Cano, 1999; Noorbakhsh et al., 2001); and (ii) the number of top educated workers over total employment, with top educated represented as the number of workers with twelve or more years of formal education (Teixeira, 2002; Wöβmann, 2003). The respondent sample presents high skill intensity (14,2% on average cf. Table 2). 7 In fact, almost half of the firms state that the number of engineers in their total employment surpassed 5% (23% said that engineers represented more than 20% of total employment). By Portuguese standards these TBFs are highly human capital intensive firms. For instance, considering data referring to 2002 from Balanço Social (DEEP, 2003), an annual compulsory survey to all 7 All variables are reported as three-year averages ( ). 8

11 firms that employ more than 100 workers, the average percentage of workers with University degree (engineers and others) was 9,9%. Similarly to the skill intensity indicator, the education intensity, which is measured by the percentage of employees with 12 years of schooling or more ( top educated ), also reflects the high human capital endowments of the firms covered in this study. Approximately 84% of TBFs pointed out that top educated workers represented more than 5% of their total workforce, with almost half of TBFs indicating that this figure exceeded 20%. For the respondent TBFs the mean of the education intensity indicator is 26,3% (cf. Table 2). Once again, comparing with Balanço Social figures, we conclude that our sample is relatively highly educated Independent variables Our strategic variable, foreign ownership, is a dummy variable which takes the value 1 in the case 50% or more of its equity is foreign-owned and 0 otherwise. The cut-off point of 50% was chosen owing to two main reasons: firstly, and without further specific information, it is the least controversial way of considering that a firm is controlled/owned by a certain type of investor, foreign or domestic; as such, it is widely used in the literature (Bellak, 2004; De Backer and Sleuwaegen, 2005), more often than the minimum threshold of 10% of capital adopted by the more controversial OECD Statistical Benchmark Definition for Foreign Direct Investment (OECD, 1999); 8 secondly, only 3% of the companies in the sample had a minority participation of the foreign investor, so the majority ownership was the main strategy when FDI occurred; 9 hence, we decided to consider majority ownership as the most accurate evidence of being a national or a foreign-owned company. Around 15% (cf. Table 2) state that foreign entities owned above 50% of the surveyed firms capital. Hence, it must be noted, that a substantial percentage of TBFs are nationally owned - 82% do not present foreign capital in their equity structure. The models estimated consider as regressors, apart from the main determinant (foreign ownership), four other structural variables (here used as controls): size, age, R&D intensity, and export intensity. 8 The OECD Benchmark Definition of Foreign Direct Investment (OECD, 1999) provides operational guidelines on how foreign direct investment (FDI) activity should be measured according to internationally agreed standards. The initial version of the Benchmark Definition (in 1983) and its subsequent revisions were prepared under the supervision of the Investment Committee. Accordingly, it is considered FDI when the proportion of firms equity owned by foreign capital is equal or above 10%. 9 Therefore, we obtained similar results when we replicated all the models with a minimum FDI threshold of 10% - this was undertaken, and the fact that the results are similar only testifies to the dominance of majorityowned FDI, when there was foreign capital in the surveyed firms equity. 9

12 Substantial literature on multinationals (mostly surveyed in Bellak, 2004) shows evidence of performance gaps between domestic-owned and foreign-owned firms. The revealed performance differences have been partly attributed to ownership, but also to firm characteristics such as size (Nickell, 1996), age, R&D intensity (Buckley and Casson, 1976; Markusen 1995), export intensity (Cohen, 1973) and other innovatory-related activities (Mowery and Sampat, 2004). Therefore, the models estimated and reported in this paper take into account these supplementary explanatory factors, defined and measured as follows. Size is proxied by the number of workers (in logarithmic form). Age is measured by the number of years in business (also in logs). The measure of R&D intensity is the ratio firm R&D expenditure divided by firm sales. This variable is similar to that used by Mohnen and Hoareau (2003) and Laursen and Salter (2004). Finally, export intensity is proxied by the ratio of exports to total sales. In addition to the five explanatory variables discussed above, 13 industry controls are included to control for different TBFs skill and education intensities across industries. Tables 1 and 2 contain more detailed descriptive statistics regarding our sample. Table 1 focuses on the surveyed firms human capital intensity, and foreign ownership, providing statistics disaggregated by industry where considerable differences in human capital intensity and foreign ownership incidence across industries can be easily observed. Table 2 provides descriptive statistics on all variables used in the model, and presents the respective correlation matrix. Table 1: Across industry variation of human capital intensity and foreign ownership, Human capital intensity Industry %workers with 12 or % foreign owned %engineers more years of formal TBFs in the industry education Agriculture, fishery and extractive industry 19,3 22,7 20,0 Food, drink and tobacco 5,6 20,0 26,1 Textiles 3,5 10,0 5,3 Wood, paper and printing 4,9 15,3 25,0 Chemicals and plastics 9,7 26,9 32,7 Non-metallic minerals 5,3 12,6 11,1 Basic metals and fabric. metal products 5,4 16,4 24,1 Machinery 8,7 17,0 10,8 Electrical 21,9 31,3 19,4 Transport and other manufacturing 8,7 20,5 20,8 Utilities and construction 11,3 22,5 8,3 Retail and wholesale 10,8 47,7 23,7 Computing, R&D and firm services 39,2 54,2 5,7 Other services 23,6 27,2 18,2 Average: all industries 15,2 27,2 14,7 10

13 Table 2: Descriptive statistics and correlation matrix Mean SD Min Max (1) Skill intensity 0,142 0, ,00 1 0,134* -0,032-0,399* -0,310* 0,368* -0,232* 0,188* 0,005 (2) Education intensity 0,263 0, ,00 1,000 0,058-0,371* -0,238* 0,212* -0,308* 0,109* -0,023 (3) Foreign owned 0,153 0, ,00 1,000 0,194* 0,052-0,097*** 0,095** 0,327* 0,685* (4) Firm size (log) 4,305 1, ,79 1,000 0,341* -0,307* 0,369* -0,018 0,128** (5) Firm age (log) 3,125 0, ,19 1,000-0,211* 0,065-0,097*** 0,090*** (6) R&D intensity 0,051 0, ,00 1,000-0,175* 0,177* -0,047 (7) Export intensity 0,265 0, ,00 1,000-0,017 0,068 (8) Foreign*R&D 0,004 0, ,50 1,000 0,273* (9) Foreign*Contacts with Universities 0,090 0, ,30 1,000 * significant at 1%; ** significant at 5%; *** significant at 10% 11

14 The majority of the TBFs surveyed (68%) employ between 10 and 250 workers, being the mean value for the sample of 74 workers. TBFs with more than 250 workers represent 21% of the total. A large percentage of TBFs are in business for a reasonable number of years (23 years approximately on average). In fact, 57% of the total sample claimed to be in business for over twenty years. Only 13% might be considered as startups (age below 10 years). 10 Overall, TBFs represented in this sample are characterised by a reasonably high intensity in R&D on average, 5,1% of their sales are devoted to R&D activities. 11 Recalling that the CIS III survey for Portugal concluded that the total expenditure in R&D activities (both intramural and extramural) by firms amounted to 0,8% of their total turnover (Bóia, 2003), we might claim that indeed our sample embraces high-technology and knowledge intensive firms. Around thirty firms (6,8% of the total) present truly remarkable average R&D intensities, above 20%. A few of these are firms whose business is totally centered in performing R&D activities. As regards markets served, the bulk of TBFs are inward oriented (exporting on average 26,5% of their total sales). Indeed, within the period covered in this study ( ), almost twothirds (63%)of the firms surveyed export less than 20% of their total sales. For Portugal as a whole, the average proportion of exports in total Gross Domestic Product in the period amounts to 30,7% (INE, 2003). The correlation matrix shows that, without controlling for other variables, skill and education intensity are negatively and significantly linearly related to size, age and export intensity, and positively (and significantly) linearly related to skill intensity. Thus, smaller, younger, exportled and technology-intensive firms tend to be more strongly associated with high levels of human capital intensity. In contrast, ownership structure fails to be linearly statistically associated with human capital variables. 10 Startup is a rather vague concept, generally meaning a new business venture in its earliest stage of development. Usually its operationalisation is made based on the age in business, ranging from 3-5 years up to 15 years. Given this wide variation, we opted for Almeida et al. s (2003) definition, which considers startups those firms with 10 or less years in business. 11 It is rather peculiar that in spite of being listed as a R&D performer, almost 20% of the respondent TBFs, when asked about the average amount spent in R&D activities in the three-year period of , declared having registered in their accounts no value for this item. Some of these firms recognised, however, to have performed R&D activities in the period in study but did not reflect these expenses in their accounts. Some of these respondents are affiliates of other firms, and stated that R&D expenses were exclusively registered in their parent firms accounts. 12

15 4. Model specification and results 4.1. Econometric model specification In order to test whether foreign ownership matters for explaining establishments human capital intensity, the following empirical models are estimated for (average values): h i = α + β f + β X + ε (1) 1 i 2 i i h i = α + β f + β ( f * R & D) + β X + ν (2) 1 i 2 i 2 4 i i h i = α + β f + β ( f * R & D) + β ( f * Univ) + β X + ω (3) 1 i 2 i 3 i 3 4 i i Where h i denotes TBF s i human capital intensity, that is, the proportion of engineers (or workers with twelve or more years of formal education) in total employment. f i is the dummy variable for foreign owned TBFs. X denotes a vector of other variables representing characteristics of the firm (size, age, R&D intensity; export intensity; and industry), likely to influence its human capital intensity. ε i, ν i, ω i are error terms following the standard assumptions. Our attention in this investigation is focused on the β 1, β 2, and β 3 coefficients, trying to assess whether the impact of foreign ownership and human capital intensity are statistically related and which of the effects are more relevant - the direct effect (β 1 ), or the indirect ones through the interaction with R&D intensity (β 2 ), or through the interaction with the frequency of University contacts (β 3 ). Although most of the literature (see Bellak, 2004) focuses on the direct impact of foreign ownership, we here try to highlight the relevance of its indirect impact, namely through technology competencies, translated into R&D intensity and propensity for drawing on external sources of information for innovation, such as universities. These latter interactions were particularly emphasized by the literature on innovation (e.g., Laursen and Salter, 2004; Mowery and Sampat, 2004), which stress the leverage effect that R&D and university contacts intensity have on firms human capital traits. In this sense, we would expect that the impact of foreign ownership on TBFs human capital intensity is higher the more intensive are TBFs R&D efforts (Hypothesis 2) and the more frequent are TBFs contacts with Universities (Hypothesis 3), that is ˆβ 2 and ˆβ 3 positive and statistically significant. 13

16 4.2. Results We find foreign ownership significant in explaining the human capital intensity of TBFs since the parameter estimates are significant for this variable. It is interesting to note that foreign ownership directly impacts on education intensity of TBFs, whereas in TBFs skill intensity the relevant impact is the indirect one, through R&D efforts. This seems to give reason to both the literature on multinationals, which emphasizes mainly the positive effects that multinationality brings to countries general levels of human capital, and the literature on innovation that highlights the mediating role of R&D on firms-industry specific human capital upgrading. This further highlights the importance of the two faces of foreign ownership foreign ownership s (indirect) impact on human capital intensity is closely related to the generation of new knowledge within the firm (Cohen and Levinthal, 1989), and the ability to seek for external sources of information and knowledge for innovation activities (Laursen and Salter, 2004; Costa and Teixeira, 2005). In this case, and as expected from Hypotheses 2 and 3, the impact of foreign ownership on TBFs human capital intensity is higher the more intensive are TBFs R&D efforts and the more frequent are TBFs contacts with Universities. Our results confirm the importance of controlling for R&D intensity and university contacts when trying to explain the firms human capital intensity. Concerning R&D intensity per se, the results show that R&D efforts are significant in explaining TBFs specific human capital (skills) intensity, but not general human capital (education) intensity. Regarding the control variables, size proved to be negatively signed and significant in all models estimated (for both human capital proxies). This means that smaller TBFs tend to be more intensive in human capital. As regards the variable age, it turned out that younger firms tend to be more human capital intensive, although in this latter case the results hold only for the models with human capital intensity measured through the variable skill intensity. These two results are not surprising as, after all, we are dealing only with technology-based firms. Literature on this type of firms (e.g., Bartel and Lichtenberg, 1987) usually highlights that smaller and younger firms tend to be more intensive in (mainly specific) human capital. 14

17 Table 3: OLS estimation, explaining the human capital intensity of TBFs located in Portugal, Dependent variable: skill intensity Dependent variable: education intensity Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Coef. p-value Coef. p-value Coef. p-value Coef. p-value Coef. p-value Coef. p-value Foreign 0,036 0,134 0,013 0,617-0,013 0,699 0,081 0,014 0,072 0,039 0,113 0,014 Size (log) -0,022 0,002-0,022 0,001-0,022 0,002-0,033 0,001-0,033 0,001-0,034 0,001 Age (log) -0,025 0,031-0,023 0,044-0,024 0,036 R&D intensity 0,190 0,009 0,156 0,033 0,152 0,037-0,022 0,156-0,022 0,169-0,020 0,209 0,016 0,874 0,004 0,970 0,010 0,920 Export intensity 0,000 0,457 0,000 0,440 0,000 0,448-0,001 0,191-0,001 0,194-0,001 0,200 Foreign*R&D 0,809 0,008 0,773 0,012 Foreign*University contacts 0,054 0,081 0,286 0,502 0,347 0,416-0,076 0,140 Constant 0,356 0,000 0,351 0,000 0,356 0,000 0,672 0,000 0,670 0,000 0,662 0,000 Foreign ownership estimated impact on TBFs human capital intensity 1 0,036 0,054 0,060 0,081 0,087 0,084 N F-statistic 13, ,000 13,351 0,000 12,728 0,000 8,296 0,000 7,871 0,000 7,516 0,000 Adjusted R 2 0,372 0,382 0,383 0,258 0,257 0,257 Note: All models have (13) industry controls. 1 HC = β 1 + β2 ( R & Dint) + β3( UnivCont) Foreign. 15

18 Export intensity fails to be a significant determinant of human capital intensity, regardless the proxy for human capital intensity. Very often (and now thinking specifically of the Portuguese case), firms that are focused on exporting most of their output do not develop high value-added activities (Tavares and Young, 2002), therefore do not display a considerable human capital intensity. Overall, and summarising the results obtained, three main outcomes of our empirical study should be strongly emphasised: 1) foreign ownership directly (and significantly) impacts on firms general human capital; 2) foreign ownership indirectly (and significantly) impacts on firms specific human capital; 3) the total impact of foreign ownership on firms human capital intensity is higher for education- (general) than for skills- (specific) related human capital intensity. We estimate that, all other factors remaining constant, 1 percentage point (pp) increase in foreign ownership tends to lead to 3,6pp - 6,0pp increase in TBFs skill intensity and to 8,1pp - 8,4pp increase in TBFs education intensity. 5. Conclusions and Policy Implications 5.1. Concluding Remarks This paper aimed at contributing to the relative scarce empirical literature studying the impact of FDI on human capital intensity. Our approach was to establish whether foreign ownership was a significant determinant of the human capital intensity in TBFs. We used the empirical setting of Portugal, a country that has been encouraging FDI inflows (lately more proactively, through various and sizeable FDI incentives) and at the same time a country with a recognised deficit in qualifications, and with some of the poorest education indicators in Europe and in the developed world. Portugal s sluggish economic growth, prevalence of low value-added activities, challenges as a FDI host economy, and relatively low stock of human capital make this study timely, by tackling these critical issues for the country s development. Using brand new evidence gathered through a purposefully-designed and representative largescale survey of TBFs located in Portugal (with a usable sample of 475 firms, 61% response rate to the comprehensive survey undertaken), we found that foreign ownership had a positive relationship with the human capital intensity of these TBFs, directly, and indirectly when foreign ownership was considered jointly with R&D intensity and the frequency of contacts with Universities. As mentioned above, three key results were obtained. Firstly, foreign 16

19 ownership directly (and significantly) impacts on firms general human capital (measured by the proxy related to education proportion of workers with more than 12 years of studies in the total workforce ); secondly, foreign ownership indirectly (and significantly) impacts on firms specific human capital, the latter measured by the proportion of engineers in the total labour force a conventional proxy for top skills ; thirdly, the total impact of foreign ownership on firms human capital intensity is higher for education- (general) than for skills- (specific) related human capital intensity. Two interaction terms were considered: one, relating foreign ownership to R&D; the other, connecting foreign ownership to the frequency of university contacts. When the dependent variable (human capital intensity) was measured by the top skills indicator, these interaction terms were significant. In particular, the significance of the interaction term foreign ownership jointly with R&D needs to be emphasised, meaning that the joint interaction of foreign ownership with R&D intensity is a strong predictor of high human capital intensity. Hence, the results provide broad support to the three hypotheses specified above when human capital refers to skill intensity (specific human capital) and also support a clear positive direct impact of foreign ownership on general human capital Policy Implications These findings have, in our opinion, relevant policy implications, both for policies aimed at attracting (and maintaining) FDI, and to broader (more macro), horizontal policies (such as those related to education and training of human resources). First of all, foreign ownership matters. Foreign subsidiaries operating in Portugal are, generally speaking, better endowed in terms of human capital. Hence, for a country like Portugal, needing badly to catch up in terms of qualification indicators (although that is a long term pursuit), with sluggish growth, a difficult external image, and that has lost its traditional labour cost advantage, our results seem to suggest that FDI should be supported. However, not all types of multinational operations. Since we are dealing only with TBFs, i.e. the most innovative firms in the country, it must be concluded that FDI-related policy ought to be discerning and emphasise the quest for higher value-added MNE activities, in a way that is compatible with the endogenous resources and capabilities that the country can realistically offer. This means adopting a selective, targeted and sophisticated approach, and a clear strategy on what kind of firm is desirable. Proactive chase for high value-added investors will, nevertheless, fall apart if it is not coordinated with other policy measures, we argue, of 17

20 broader nature. For instance, altering the supply of tertiary (and even secondary and technical) education courses in line with the strategy delineated (e.g. Ireland increased dramatically the number of courses offered to engineers, particularly in electronics, in line with its fine-tuned industrial targeting; Costa Rica attracted Intel, more due to its commitment to changing secondary education curricula to emphasise electronics and English; and many other examples exist, like the best practice case of the Penang Skill Development Centre in Malaysia, as well as other interesting experiments in countries like Singapore and Thailand). After all, each country gets the investment she deserves. FDI policies, especially untied subsidies and tax cuts, cannot make miracles: if they are too easy, and given without a proper strategy, they can even be counterproductive and foreign investors know how to assess the other side of the bargain. Probably the way to go would be to tie incentives to high valueadded operations, in a transparent and clear way so that the investor knows what he can get depending on the investment s characteristics. This would weed out purely opportunistic, rent-seeking and incentive-snatching investors, and would give a more serious sign that the country really is committed to changing its model of development, and its way of acting. A focus in some niches where excellence has been achieved (e.g. in microelectronics, with the Siemens case) should be eventually pondered, and eventually to support more FDI in specialised services (that is a more sensitive issue, but on average, tradable service activities are o average more intensive in human capital than the average industrial operation). A balanced, and more systemic approach is needed, encompassing a more coordinated implementation of measures, and, above all, an emphasis on the quality of operations, in order to attract sustainable MNE operations. Energy and resources should be devoted to investors that are already present in the country often, it is harder to maintain than to attract investment. But with a closer relationship with subsidiary managers of the truly high quality operations, giving less attention and not wasting resources with those subsidiaries that are not bound to be upgraded. This would imply a deep knowledge, and very accurate data on the country s foreign investor population, with objective (and again transparent) criteria. Notwithstanding, FDI-related policies are unlikely to be enough. In terms of the usual investment in education and training, it is urgent that a better allocation of resources is achieved, as Portugal per capita spending on education (the input measure) 18

21 is relatively high, but the output indicators are simply not in tune with the relatively high spending so, investing better, and not necessarily more, as a priority. However, the main hindrance to a credible image of the country vis-à-vis foreign investors is the remarkable institutional and policy instability that is obvious to the eye of any investor. FDI-related policies have to be implemented by a stable, independent, uncontroversial, and credible institution, that has the legitimacy and the ability to decide on most of the issues that interest an investor (foreign or national). This paper s results, are, in sum, in line with general expectation, but are specifically interesting given the direct and indirect impact of foreign ownership according to the human capital measure. It seems, thus, that FDI is acting in the right way impacting directly and positively on general human capital, which, in terms of spillover potential (positive externalities) to the rest of the economy, is more important as these general qualifications are easily transferable across companies and sectors. Therefore, foreign multinationals operating in Portugal are helping with the development process, and also seem to offer some hope that they have invested in Portugal not only to take advantage of low costs per se, a selfobsolescing (and already eroded) source of competitive advantage. In a nutshell, both FDI and human capital formation should be promoted, as two relevant key engines of growth and also of sustainability of the Portuguese economy, trying to promote properly a consistent and well communicated strategy, acting also in order to shape the skill mix towards more engineering/technical graduates, acting more strongly on intermediate levels of education, on vocational/technical training, continuing education, on knowledge infrastructure policies, on linkages between the foreign and domestic agents. Catching up in terms of human capital stock will take decades. However, starting to act upon it cannot be delayed any longer, if the low-income, low-skill trap is to be avoided. Acknowledgements The authors are deeply indebt to Joana Costa for the gathering of the data used in the present work. 19